Global firms and wages:
is there a rent sharing channel?
Jože P. Damijan and Luca Marcolin
Summary
In this paper we explore the scope of international profit sharing of firms engaged in international trade and FDI
activities. We extend the approach proposed in the literature by allowing not only for differences between domestic
and foreign owned firms, but also between firms with outward FDI, importers and exporters. We argue that firms
engaged in international trade enhance their performance through knowledge spillovers and technology upgrading
similarly to what happens for companies which are part of multinational groups, and that this superior performance
can translate into substantial wage premia to workers through profitability. Using a unique dataset for Slovenian
firms for the period 1994-2002, we confirm the existence of positive profit sharing for foreign firms, firms that
engage in outward FDI and two-way traders, but not for firms which only import or export.
Keywords: firm heterogeneity, multinational firms, wage determination, profit-sharing.
JEL Classification: F23, J31, J50.
Contact details:
Joze P. Damijan, Faculty of Business and Economics, University of Ljubljana. Kardeljeva pl. 17, 1000 Ljubljana,
Slovenia. Phone: +386 (0)1 5892 763 ; email: [email protected].
Luca Marcolin, Faculty of Economics and Business, KU Leuven. Naamsestraat 61, 3000 Leuven, Belgium.
Phone: +32 (0)16 323798 ; email: [email protected].
Joze P. Damijan is Full Professor of Economics at the University of Ljubljana, and a research affiliate at IER,
LICOS and VIVES (KU Leuven). Luca Marcolin is a PhD candidate in Economics at KU Leuven.
This paper has benefitted from discussions with the participants of the ETSG 2012, CEPR & GIST 2012, and LETC 2013 conferences, and to
seminars at the EBRD, KU Leuven, and University of Ljubljana. We gratefully acknowledge their contributions.
The working paper series has been produced to stimulate debate on the economic transformation of central and
eastern Europe and the CIS. Views presented are those of the authors and not necessarily of the EBRD.
Working Paper No. 164
Prepared in December 2013
Introduction
Over the course of the last three decades, globalisation characterised by free movements of capital
and goods has changed the shape of the world economy. Trade theory justifies the necessity for
free movements of goods between countries on the account of the gains from trade. These gains
may arise from an improvement in the terms of trade, better utilisation of resources through
specialisation along the comparative advantage or, more recently, from an increase in the number
of varieties of goods available in the country and a decrease of prices on average. At the same
time, ever since the statement of the Stolper-Samuelson theorem1 , it is clear that these gains
may be unevenly distributed across production factors, so that the global positive welfare effect
of trade may hide important adjustments on the use of and return to factors. This phenomenon
is magnified by the emergence of movement in factors, which is embodied by multinational
corporations. Recent evolutions in trade theory and in data availability have allowed discussion
and testing of the existence of heterogeneous responses of factors to internationalisation not only
across sectors, but also across firms and plants.
In this paper we aim to contribute to the growing literature linking globalisation with changes
in the return to the labour factor. Economic research on wage premia of internationalisation is
quite extensive and encompasses the response of wages to both foreign ownership and trade in
light of their complementarities. Studies show that foreign affiliates perform better than domestic
firms on a number of parameters including wages.2 Almeida (2007), in an insightful review of
the literature, points out that these foreign wage premia can be motivated by the increase in
productivity and skill level achieved thanks to the utilisation of a better foreign technology, or
because foreign firms, lacking the ability to attract the best workers, must offer higher wages
than domestic firms for the same quality of workers. A similar effect is noted for importing and
exporting firms, suggesting that firms need a higher average skill content of the workforce when
deciding to serve foreign markets and exploit imported inputs of higher technological content.
Similar phenomena have been discussed by others in the literature.3
Our paper, however, does not focus on the wage premium related to the change in firm skill mix
imposed by internationalisation, but instead explores the second channel of profit sharing. According to this literature, a firm making positive profits may be required to share them with its
employees if the bargaining power of the latter is strong enough. For a given level of profits,
the higher the bargaining power, the more important the profit share employees can extract from
employers. On the other hand, it is also true in this framework that for a given bargaining power
of employees, more profitable firms will offer higher wages. We follow on this path the sem1 The
SS theorem explains how a change in the price of goods due to international trade affects the prices of
factors used in the production of such goods. As factor endowments differ between countries, international trade
impacts the return to labour and capital differently in different countries. More recently the literature has proved the
robustness of these conclusions to the relaxation of some assumptions, in particular the number of goods and factors
used in production.
2 We refer to Aitken et al. (1996); Feenstra and Hanson (1997); Sjholm and Lipsey (2006); Girma and Gorg
(2007); Matthias Arnold and Javorcik (2009).
3 See, for instance, Bernard and Jensen (1997) and, more recently, Goldberg and Pavcnik (2007) in their review
of literature on inequality, Irarrazabal et al. (2009) for Norway, and Hanson and Harrison (1995), Verhoogen (2008)
and Frias et al. (2012) for Mexico
1
inal papers of Abowd and Lemieux (1993) for Canada; Blanchflower et al. (1996) for the US;
Van Reenen (1996) for the UK; Margolis and Salvanes (2001) for Norway and France. We will
offer a stylised theoretical model trying to describe the rent sharing behaviour.
This paper contributes to the debate by taking into account different modes of internationalisation
in an unified approach, and linking them to rent sharing. In particular, we are interested in
pointing out the differences in rent sharing behaviour by trading and ownership status of firms.
We derive the intuition of our empirical specification from a simple model of profit sharing, and
we interpret our results in terms of changes in relative bargaining power of employees due to
the firm internationalisation process. In particular, we ask whether trading firms consistently
differ in terms of profit sharing from firms that only transfer part of their production abroad.
Moreover, we explore the degree of profit sharing for firms that do both activities. These aspects
have so far been ignored by the relevant literature, which, to our knowledge, has only recently
investigated the rent sharing behaviour of exporters in theory (Cosar et al. (2010); Helpman et al.
(2010, 2011)) and empirics (Macis and Schivardi (2012)). As far as the rent sharing behaviour of
multinationals is concerned, we refer to the works by Budd et al. (2005) for the European context
and Martins and Yang (2010) for 47 world countries.
To assess the impact of internationalisation of firms on rent sharing, we make use of a unique
dataset for Slovenian firms for the period 1994-2002. Slovenia was chosen both for the high
quality and availability of the panel data on trade and foreign investment at the firm level and its
extensive trade openness. To quantify the effects of the firm international status, we investigate
the evolution of wages at the firm level conditional on firm skill composition and other determinants of wages. Our main specification controls for possible omitted variable biases using firm
and time fixed effects. Out of concerns for possible endogeneity of profits and wages, we also
estimate the empirical model with System-GMM, thus exploiting lags of the independent variables as instruments for the variables themselves. We confirm that profit sharing in Slovenian
firms is more important for foreign firms and firms engaged in outward foreign direct investment
than for domestic firms. This suggests that employees in companies which are part of multinational groups (as either parent or affiliate) are able to extract a greater share of profits, thus
benefiting from the greater product-market power of the company. Two-way traders engaged in
both importing and exporting also display a higher degree of rent sharing. Being part of an international supply chain obliges firms to offer higher wages (through profits) in order to keep the
production process undisrupted. An interruption in the production within a GSC can have costly
repercussions on other stages of the production outside the country of reference, which increases
the workers’ bargaining power with respect to their employers. Lastly, our best specification suggests that importing only firms also offer higher wages than the others through the rent sharing
mechanism while exporting only firms do not.
Our paper contributes to two main streams of literature. On the one hand, it confirms the existence
of wage premium for internationalising firms and, on the other, it provides further evidence in
support of the profit sharing theory in an international context.
The first and second sections of the paper contain our theoretical model and its empirical derivation, as well as the empirical problems which we will take into consideration in the analysis.
The third section presents the data and some descriptive statistics. The results of the empirical
2
estimations and robustness checks are discussed in the fourth and fifth section. Our concluding
remarks follow.
1
Theoretical background
In a standard competitive labour market model, differences in wages across firms and industries
(besides differences in working conditions which need to be compensated) do not exist. As soon
as the assumption of perfect competition in the labour market is relaxed, however, wage differentials have been explained through several mechanisms, such as efficiency wages or models of
“fairness” in the labour market, for instance. We present here a very simple model where wages
will be determined by a bargaining process between workers and employers. Wages in an equilibrium with frictions will lie in between the marginal product of labour (which would reduce firm’s
profits to zero, as in the perfectly competitive labour markets) and the outside option for the workers (that is to say, their reservation wage), depending on the relative bargaining power of workers
and employers. We believe that this setting is consistent with the labour market institutions in
Slovenia and in developed countries in general, where wages are negotiated between unions and
the employers. In Slovenia negotiations take place at national, sectoral and firm level. National
bargaining focuses on differences between public and private employees, while the sectoral level
adjusts wages on the basis of educational and effort requirements of each occupation in the sector.
Lastly, negotiations at the firm level can adjust the salary upwards with respect to the provisions
in the industry.4 Union density is approximately 50 percent, but the coverage is estimated to 90
percent, which makes bargaining an interesting channel to explore for the understanding of wage
dynamics in Slovenia.
We hence follow Blanchflower et al. (1996), Margolis and Salvanes (2001), and Abraham et al.
(2009), and assume that wages are determined at the firm level through a Nash bargaining process
in which unions maximise the expected gains of their members:
u (w, w) = u(w) − u(w)
(1)
Where u is the concave utility function of individual workers or of unions, w is the workers’
salary and w is the reservation wage of workers in case the bargaining process fails. Firms, on
the other hand, maximise their profit function:
π = p f (n, k) − wn − rk
(2)
4 In
1997 a labour market reform removed the limitations to firm upward salary adjustments, as well as the
possibility of downward adjustments for firms displaying negative performance.
3
Where p is the price of output, n is employment, k is capital, f (n, k) is the production function
concave in n, r the return to capital . The outside option for the employer is assumed to be zero
profits. The solution to the Nash bargaining process maximises:
n
o
φ
1−φ
argmax(w,n) [u (w, w) n] [π]
(3)
Where φ is the parameter expressing the bargaining power of the worker and has value between
0 and 1. The first order conditions for wage and employment read, respectively:
u0 (w)φ
1−φ
−
=0
[u(w) − u(w)] n
π
(4)
φ (1 − φ) [ f 0 (n) − w]
+
=0
n
π
(5)
Re-elaborating equation(4) and using the approximation u(w) − u(w) ∼
= (w − w)(u0 (w)), one obtains:
w∼
= w+
φ π
1−φ n
(6)
or, more generally,
w∼
= c(w0 , Z) +
φ π
1−φ n
(7)
The equation displays the three fundamental channels affecting the bargaining process (Kramarz
(2003)): the firm financial condition, captured by its profitability; the bargaining power of employees relative to employers φ; the threat point by workers in the bargaining process, i.e. the
industry wage w0 which they would obtain in case of failure of the wage negotiations, as determined by a number of conditions (Z) at the institutional or company level affecting both w0 and
the bargaining process. In the current specification, capital, not entering the utility function of
unions or workers, does not impact the outcome of the bargaining process, hence it is not present
in equation(7).
4
2
Empirical model and econometric issues
Equation(7) in the simple model described in the previous paragraph suggests the following empirical specification:
wit = w + β1
πit
+ β2 Zit + εit
nit
(8)
where wit is firm’s i wage at time t, and w is the reservation wage (w0 in the theoretical model, but
φ
it needs not always be the industry wage); β1 = 1−φ
is the relative bargaining power of workers
πit
which is assumed constant in time, nit the firm’s profitability per employee in the firms, Z a vector
of regressors affecting the company’s wages, and εit an error term. Most articles linking the
wage bargaining process with rent sharing (for instance Margolis and Salvanes (2001); Martins
and Yang (2010)) test this baseline specification. In this paper we augment equation(8) with
dummies indicating foreign ownership, outward FDIs, export and import decisions of the firm,
and interacting them with the profit per employee variable.5 We hence estimate:
wit = w + β1
πit
πit
+ β2 Dsit + β3 Dsit + β4 Zit + εit
nit
nit
(9)
Where D stands for an indicator (dummy) variable, and superscript s stands for foreign ownership, outward FDI, import only, export only, both exporting and importing. β2 identifies the
straight-forward wage premium to the internationalisation mode, but we are most interested in
the sign of β3 . We expect it to be significantly different for different dummies, thus identifying
differences in rent sharing for different types of internationalised firms. In higher specifications,
we also include more than one dummy variable and interaction effect at the same time:
wit = w + β1
πit
πit
πit
+ +β2 Dsit + β3 Drit + β4 Dsit + β5 Drit + β6 Zit + εit
nit
nit
nit
(10)
with s 6= r = foreign ownership, outward FDI, export only, import only, twoway trade. In order
to be consistent with the findings in the theoretical model, we estimate our baseline equations
using both wages and profits in levels, so that β1 is the euro-increase in wages when the firm
earns an extra euro of profits.6 In our estimations Zit includes mainly: i) the firm-specific ratio of
capital to labour (fixed assets over employment); ii) a set of industry and time-specific dummies
and iii) in higher specifications, a set of unobservable company-specific fixed effects to capture
5 The
rationale for doing this is explained in greater detail in Section 4.
view of the number of observations displaying strictly negative profits (33.6 percent of all available observations), taking the logarithm of profits would censor our sample and bias the coefficient for the main independent
variable of interest. We therefore exclude this possibility and include profits in level only.
6 In
5
firm unobserved heterogeneity which might affect wages. We include i) as an proxy for the
firm’s labour skill composition, in light of the positive correlation between capital intensity and
skill level of the labour force (Griliches (1969); Duffy et al. (2004); Yasar and Morrison Paul
(2008)). Despite the imperfection of this proxy for skill composition, omitting it would bias our
results if one believes that changes in profits affect wages through changes in the composition
of the firm’s workforce. The impact of skill composition on the average salary would not be
captured by the dummies for internationalisation modes, nor by productivity, which we include
in our robustness estimation in the form of total factor productivity (TFP). Time fixed effects
are included in order to control for unobserved characteristics which are in common to all firms
but varying over time, such as business-cycle related phenomena (economic trend or shocks,
skill biased technical change). We also include industry- and firm- fixed effects to control for
unobserved heterogeneity such as firm efficiency, workers features, or preferences of unions and
workers over the reservation wages. Last, we correct for heteroskedasticity and serial correlation
by clustering the errors at the firm level and by netting the error from their AR(1) component.7
With regards to the potential endogeneity of profitability to wages, it is possible that firms solicit
effort from workers offering efficiency wages, so that wages determine profits and not the vice
versa. In this case, the coefficient on profits which we would estimate with OLS would be biased
downwards. A shock in product demand may also affect the marginal product of labour and firm
profitability simultaneously. A third form of endogeneity is introduced by measurement error,
which can be especially sizeable in this context, where profits per employee and wages are both
retrieved from the balance sheet records of the firm. In presence of a fixed effect estimation,
measurement error would bias the estimated coefficient for profit per employee towards zero,
especially if variables tend to be persistent and data are differenced (Griliches and Hausman
(1986)). We address these three sources of bias by estimating equation(10) in differences through
GMM. Once the fixed effect is netted out, and in absence of first order serial correlation in errors,
we can exploit appropriately lagged values of the regressor (here: profit per employee starting
from time t − 2 and earlier) as instruments for the regressor itself in the equation in differences
(Arellano and Bond (1991); Arellano and Bover (1995); Blundell and Bond (1998)); the standard
errors are obtained using the Windmeijer (2005) correction for finite samples. We check for the
joint validity of the chosen instruments and the absence of second order serial correlation in errors
by looking, respectively, at the Hansen and Arellano Bond tests. The endogeneity of profit per
employee makes the cross products between these and the dummies for international status of the
firm endogenous as well. What is more, the dummies themselves are endogenous, if the error of
equation(10) contains a phenomenon which simultaneously impacts the level of wages and the
international status of the firm. One such example is workers’ ability, which is reflected in the
worker’s salary and is also most likely taken into account by foreign companies when deciding
7 We
test the presence of autocorrelation in the error in two ways: a) we use the user-written command xtserial
which performs a Wooldridge F test for autocorrelation in panel data and b) we run an OLS regression of wages on
profit per employee and industry and time dummies, we retrieve the estimate of the residuals and we run a second
regression of the residual over the lagged residual, wages, profit per employee and dummies. The F test for the
equality to zero of the coefficient for the lagged residual (positive and significant) is rejected. This is unsurprising,
especially in view of the usual persistence of wages from year to year (Hildreth and Oswald (1997)). The correlation
of current and lagged wages in our panel varies from 0.70 to 0.81, depending on the year of observation. The
correction of the errors from their AR(1) component in the panel setting is achieved using the Stata command xtregar
and the option “fixed effect”. This command first transforms the data so as to eliminate the AR(1) component of the
error term, and then transforms it again (by taking differences) to purge the fixed effect.
6
to invest in Slovenia8 , but which is not observed by the researcher. As a consequence, dummies
and cross products are also instrumented by taking appropriate lags.
3
Data and descriptive statistics
3.1
Data
In this paper we exploit an unbalanced panel of Slovenian manufacturing firms for the years 19942002. The database contains information on firm balance sheet collected by AJPES (Agency
of the Republic of Slovenia for Public Legal Records and Related Services), which was then
matched with data on trade and FDI flows by firm, as collected by the Statistical Office of the Republic of Slovenia. Data are expressed in euros which we deflated using the Slovenian producer
price index at 2 digit level for value added, profits, wages and sales, and the consumer price index for capital stock. We also experimented with the gross output price index and the gross fixed
capital formation price index for Slovenia from the EUKLEMS database, which leave the results
substantially unchanged.
In the considered years, no minimum threshold for the reporting of trading transactions was
in place and therefore the database contains the sum of all trade flows involving the observed
company. We know if the firm has one or more affiliates abroad and where these affiliates are
located, and if the firm is foreignly owned, but we do not observe the share of foreign control.9
By the same token, the dataset does not contain hourly wages, hence we construct our main
dependent variable by dividing the total firm-level annual wage bill by the number of employees
working for the firm10 . Lastly, in our baseline estimations we exploit the accounting measure of
profits which is reported in the dataset, but we also experiment with profits defined as the value of
sales from which material costs and total wage bill were subtracted, a concept which intuitively
corresponds to the rents to be shared among workers (Blanchflower et al. (1996), Budd et al.
(2005)).11 Our results are similar using either of the two measures.12
8 The tendency of MNEs to buy the best performing domestic companies is well known in the literature exploring
wage premia of foreign ownership, and goes by the name of “cherry picking”. Read, for instance Matthias Arnold
and Javorcik (2009).
9 This was found to be of some importance in previous studies, such as Budd et al. (2005).
10 Hourly wages are preferred in the profit sharing literature, since they are more likely to be the object of negotiations between worker and employer, and they suffer less from noise in the measurement due to non-salary worker
compensations. On the other hand, as long as this measurement error is constant in time, taking equation(10) in
difference should milden the problem.
11 Reported profits may in fact be an unreliable measure of true firm profitability, due to the firms discretionary
power in reporting profits in different years.
12 Tables available upon request.
7
3.2
Descriptive statistics
We clean our dataset by discarding firms with negative values for sales, tangible assets, employment and material costs, but allowing both negative value added and profits. We are then left
with 37,261 firm-year observations, corresponding to 3,542 to 4,224 firms operating each year
between 1994 and 2002. These represent on average 22 percent of total Slovenian firms operating
in manufacturing, but also 86 percent of the country’s employment, total wage bill and revenues
from sales in the same sectors.13 These companies are heterogeneous in wages and profits, depending on both ownership and trade status. Figure 1 displays the distribution of firms by deflated
profits in our sample: foreign firms displaying average profits are much fewer than the domestic
ones while the distribution for this category of companies is much more dispersed. Firms which
both import and export also have a more dispersed distribution than firms which only import or
export. Similar differences are found when looking at average (deflated) wages: the distribution
for foreign owned firms is moved to the right of that for domestic firms, as is the distribution of
two-way trading firms with respect to one-way trading ones.
The evolution in time for profits and wages is also different for firms of different ownership and
trading type, as displayed in Figure 2. Foreign and two-way trading firms earn consistently higher
profits than domestic and one-way trading firms, and the difference is increasing in time.14 The
same ranking is found for average wages: if foreign firms offer consistently higher wages than
domestic firms, so do two-way traders with respect to firms engaged in import or export only.
Exploring the rent sharing behaviour by type of firm may help understanding such heterogeneity
in profits and wages.
Further differences among these types of firms are summarised in Table 1 and 2. In line with the
literature, foreign firms are on average bigger and more capital intensive than domestic firms; they
sell more and produce a higher total value added. Foreign firms which only export represent an
exception, as they are on average smaller, less capital intensive and profitable than the domestic
ones. Firms which only export are in fact smaller and worse performing even than the average
firm, no matter what the ownership status. In comparison, firms which only import do better in
all respects, though they still report one tenth of the sales and value added than two-way traders.
It is therefore clear that two-way traders perform better than one-way traders, while the picture is
less clear-cut when it comes to the ownership of the firm, with foreign companies performing on
average better than domestic ones.
4
Empirical hypotheses and results
This section presents our main results. We first report the specifications where the issue of potential endogeneity is ignored, and then move to the System GMM estimates. Lastly, we describe
13 Data
for aggregate turnover, cost of personnel, employment, number of firms in the concerned years are taken
from Eurostat and the Statistical Office of the Republic of Slovenia.
14 It should not surprise that two-way trading firms in 1995 made negative profits on average, as in 1995 the
aggregate profits of all Slovenian firms were negative, a result of transition recession.
8
the results of our robustness checks. For each result we present the implied wage elasticity to
profits, and the so called Lester range, after Lester (1952). This is defined as the elasticity of
wages with respect to profits multiplied by four times the ratio between the standard deviation of
profits and mean profits. It is interpreted as the wage increase a worker would experience if she
moved from a firm with profits two standard deviations below the mean to a firm with profits two
standard deviations above the mean, everything else held constant.
4.1
Ownership status - OLS and FE
We first estimate equation(10) with OLS and firm-level fixed effect, where we include the dummies for foreign ownership and outward FDI. If profit sharing truly takes place, we expect the
coefficient on profits per employee to be generally significant and positive. We also hypothesise that the coefficient on the interaction term between profits per employee and the dummy
for foreign ownership is significantly different from the one on the average profit per worker.
Rodrik (1997) pointed out that if workers in the domestic and foreign country are substitute for
one another, the bargaining power of domestic workers is lower for more international firms. On
the other hand, Budd et al. (2005) find a positive impact of parent profitability on the affiliate’s
wages, and vice versa, that is to say evidence of international profit sharing. A positive sign in the
cross product of profits and the foreign dummy would imply that workers are able to appropriate
a share of the profits of the international group. This is especially true if the relevant product
market for wage setting goes beyond the national borders as in Budd and Slaughter (2004). We
remain similarly agnostic on the sign of the coefficient on the dummy for outward FDI: a positive
sign would be further proof of the international rent sharing behaviour described by Budd et al.
(2005), for which the gains in efficiency created by the parent while offshoring translate into
higher profitability; a negative sign would reflect the higher bargaining power of employers in
the headquarters, who can threaten to move away further parts of the production.
Table 3 contains our first answer to these questions. The coefficients on Π/N (that is to say, profit
per employee) suggest that higher profits are shared with the employees and this has a positive
impact on wages. This is true even if we simultaneously control for our proxy of skill content
of the firm, or the ratio of capital to the number of employees (K/N). The coefficient on this
variable is in itself positively and significantly correlated with wages, since a higher average level
of capital in the firm is reflected in a higher average wage. However, the baseline specification
in columns 1 and 4 ignores the issue of potential heterogeneity in the rent sharing behaviour due
to the existence of different types of firms, and the impact of such diversity on wages. Columns
2 and 5 investigate the possibility of differences in profit sharing between domestic and foreign
firms. We find confirmation of the existence of a wage premium of foreign ownership, as the
coefficient on the foreign ownership dummy (D f ) is always positive and significant, although at
different levels across specifications. The coefficient on our variable of interest, the cross product
between the dummy for foreign ownership and profit per employee (D f ∗ Π/N), is positive and
significant, suggesting that a foreign company shares its profits more than a domestic firm in
the form of higher salaries for its workers. The elasticity of wage to profits is also positive and
greater for foreign firms than the average firm, ranging from 0.0006 to 0.0022.
9
Unsurprisingly, coefficients estimated by OLS on both profit sharing and profit sharing of foreign
companies (columns 1 to 3) are systematically biased upwards, as OLS suffers from omitted
variable bias despite our attempt to control for industry and time fixed effects. When applying
the firm level fixed effects specifications (columns 4 to 6), we only exploit the information carried
by firms switching status, and the coefficients on both profits per employee and the cross product
decrease up to one third of the OLS ones.15 A firm switching from domestic to foreign ownership
increases its average wage by 0.03 euros for each extra euro of profits. Our estimates are in line
with those reported in the international rent sharing literature, with a Lester range associated to
foreign ownership going from 18 to 26 percent.16
Our results do not change qualitatively when we further include dummies for outward FDI
(columns 3 and 6). Although the elasticity of wages to profits for foreign firms (and, accordingly, the Lester range) slightly decreases in magnitude, the sign and significance persist. A
Slovenian firm which engages in outward FDI is found to offer a wage premium to its workers
through profit sharing. This evidence is coherent with the idea that having an affiliate abroad
does not decrease the bargaining power of workers, and that on the contrary the latter benefit of
higher infra-group profitability.
4.2
Trade status - OLS and FE
We now turn to the examination of the rent sharing behaviour of trading firm. We first estimate
equation(10) with dummies indicating that the firm is importing only (Di or Imp), exporting only
(De or Exp) or importing and exporting at the same time (Db or Both). We might expect the sign
on the import-only cross-product to be negative: a firm engaged in importing may more easily
substitute away its employees with imported inputs, thus undermining the bargaining power of
the workers. On the other hand, if the number of employees and the skill ratio in the firm does not
change when starting to import, the number of imported variety and their quality may require the
firm’s workers to exert more effort, which would translate into a higher relative bargaining power
of the workers. The same outcome is obtained if workers are specific to some imported inputs,
since the absence of one worker would interrupt the production chain. The expected sign on
the import-only cross product is therefore ambiguous. Brock and Dobbelaere (2006) argue that
export competition increases the tightness of the labour market, thus increasing the bargaining
power of workers relative to the company. Similarly, export competition further decreases the
bargaining power of the firm by making the latter more dependent on its workforce for efficient
and regular production. Exporters are then expected to have positive profit sharing behaviour.
Macis and Schivardi (2012) find that exporters display strong and positive rent sharing behaviour.
The hypotheses for one-way traders may also hold for two-way traders, but a priori we cannot
say which between the effect (importing and exporting) may dominate. On the other hand, the
labour substitution hypothesis stands in contrast to the evidence that these firms have ten times
the turnover and value added of a firm which is only importing, and are making on average
15 Results
are obtained using the command xtregar, in view of the time persistency of wages which we mentioned
in section 3. These estimates are also usually lower than the ones obtained simply clustering the errors, so that if we
are wrong, we are understating the importance of the rent sharing phenomenon rather than overstating it.
16 The estimates of elasticity may appear small. This statement, however, does not take into account the high
volatility of profits, which can significantly change from one year to the other (Hildreth and Oswald (1997))
10
consistently higher wages starting from 1996 onwards. Moreover, firms both importing and
exporting are likely to be part of an international supply chain, making it more costly to interrupt
production, in light of the repercussions which these interruptions may have on different stages
of the production chain. This would suggest a positive sign on the cross product for two way
trader and profit per employee.
Our preliminary answers are reported in Table 4. Signs reflect the expectations in the formulated
hypotheses: switching to exporting or to two-way trading affects the degree of rent sharing positively, as opposed to importing only. Looking at columns 7 and 8 also suggests that the elasticity
of wages to profits is higher in firms which start both exporting and importing. On the other hand,
these conclusions may be affected by the existence of firms which are trading but also part of a
multinational group. That is why in Table 5 we augment equation(10) adding indicators for both
foreign ownership and trading status of the firm.
Including further controls worsens the significance of some of the estimates in fixed effects,
which does not come as a surprise, as international activities are often correlated to one another:
a foreign firm is more likely to be engaged in outward FDI and/or trade than a domestic one, for
instance. Wages for workers in firms which start importing but not exporting are lower, the higher
the level of profits.17 The sign of the coefficient is consistent with our hypothesis of reduced bargaining power of workers when employment can be substituted away with imports. Accordingly,
both the elasticity of wages to profits per employee and the Lester range are negative when the
coefficient for Di is significant. Exports only, on the other hand, seem to generate positive rent
sharing once again (columns 2 and 4 in Table 5). Turning to the rent sharing behaviour of twoway trading firms (columns 3 and 4), the sign is positive and robust across all specifications.
While switching to both import and export at the same time, firms increase their profits relatively
to firms which are not doing so.
4.3
Treating endogeneity: GMM
Until now, our results did not take into account the possibility of measurement error in our covariates, as well as of the endogeneity of profits per employee. To address both issues we revert
to System-GMM, whose results are presented in Table 6. To dispel doubts on the validity of our
instrumenting technique and overfitting of the variable of interest, we also report the p-value of
the Hansen and Arellano-Bond tests and the number of used instruments.
Although the direct profit sharing on wages when dummies for internationalisation are included
sometimes disappears, the existence of heterogeneity in profit sharing behaviour by internationalisation mode is confirmed. All signs are the same as in the FE specification without correcting
for endogeneity for ownership and two-way trade (columns 5 and 6), but not for one-way traders
(columns 3 and 6). This evidence is consistent with the complementarity between the firm’s
workers and imported inputs, which makes the former fundamental for the continuity of the pro17 Note
that a positive significant coefficient on Di associated with a negative coefficient on Di ∗ Π/N would not
be contradictory. Even when controlling for the firm’s skill composition and rent sharing behaviour, some wage
premia may pass through other channels which are not explicitely observed.
11
duction process, and increases their bargaining power with respect to employers. The coefficient
on the cross product for one-way exporters switches sign, but it is insignificant in both specifications (columns 4 and 6), making it impossible to draw a clear inference on the underlying
phenomenon.
The magnitude of the elasticities generally increases, which is consistent with the existence of
endogeneity in the OLS and FE estimation, where the causal effect of profits is downward biased.
The same increase in coefficients is reported in the literature, when endogeneity is corrected by
either GMM (Budd et al. (2005); Martins and Yang (2010)) or IV (Abowd and Lemieux (1993);
Arai and Heyman (2009); Card et al. (2010)).
5
Robustness checks
Overall, the presented evidence suggests the existence of profit sharing for both foreign and
trading firms with respect to domestic and/or non-trading firms. In order to assess the robustness
of our results and support our claims, we present in this section other different specifications of
equation(10).
In a first set of robustness tables we experiment with different controls for firm performance (see
Table 7). We introduce, alternatively, total factor productivity estimated using Levinsohn and
Petrin (2003) algorithm (column 1) and a proxy for firm size, that is to say employment (column
2). Each regression is estimated with firm fixed effect correcting for AR(1) serial correlation in
errors. TFP is found to have a positive impact on wages, as per economic intuition: more productive firms can afford to pay higher wages, or simply more productive firms are such because
their workers are more productive and are therefore paid more (endogeneity is not controlled for
in this specification and we do not control for skill composition). Size in terms of employment is
negatively correlated with wages, as in a standard model of perfectly competitive labour market.
The sign of the interaction terms of interest are coherent with previous estimations: changing
ownership (inward and outward FDI) and two-way trade status is found to impact wages positively and so does starting to export without importing. Importing only, on the contrary, are found
to impact wages negatively. Although elasticities and Lester ranges are not computed for these
tables, the magnitude of the coefficients is comparable between with Tables 3 and 5.
In Table 8 we then restrict the analysis to firms displaying negative profits. Firms with negative
profits cannot redistribute them to workers and therefore we expect the coefficients to the cross
products between firms international status and profits to be insignificant. The estimation results
support this intuition. On the other hand, were firms able to freely reduce wages in times of operational losses, we would expect a positive and significant coefficient on the cross products. This
may not happen, however, if the capability of the firm to reduce wages is limited by institutional
provisions (such as the one mentioned in Section 1), or by sectoral agreements. These two factors
may explain the negative coefficient on the level of rent sharing in completely domestic firms:
salaries still increase even if profits do not. This result may also reflect the willingness of the firm
to keep its workers from switching company, especially if the operational loss is believed to be
12
temporary.
In order to assess the importance on our results of the change in wage policy introduced by the
Slovenian government in 1997, we estimated equation(10) separately for the period before and
after 1997. In 1997, the government reduced the social contributions rate for employees while
introducing the payroll tax. The payroll tax was differentiated according to the wage bill, starting
with zero for the lowest bracket and putting the highest rate at 14.8 per cent for the highest salary
bracket. This in turn changed the wage policy of firms shifting aggregate wages in favor of lower
wage brackets. Net effects of these policy interventions depended on skill structure of firms, but
could potentially increase profit sharing in labour-intensive industries.
The results are included in Table 9, where in the first column “POST” is an indicator having value
1 from year 1997 onwards while in column 2 it has value 1 from the following year. Neither after
the introduction of the policy (column 1) nor in case the change in policy was anticipated (column
2), are all the coefficients insignificant before or after the reform. If it is the case that rent sharing
changes sign for some firm status after 1997, the change is not always the same for all types of
firms, which is sufficient to confirm that pooling information over the entire sample is not biasing
our results in a specific direction.
Lastly, we ask whether rent sharing is not affected only by the international status, but also by
the value of international transactions. As our database contains the value of exports and imports
at the firm level, we interact the shares of trade in total revenues of the firm with the profit
per capita variable. It may indeed be the case that workers’ bargaining power with respect to
employers increases with the value of the traded goods, independently on the size of profits. This
may be the case, for instance, if interruptions of production are especially costly for the employer
who is exporting high value goods. Also, if imports are especially valuable (in relative terms to
revenues from sales) they may be embodying higher technological contents or quality. This may
require workers to input more effort in the production (given a fixed skill content of labour in the
firm in the short run) which would translate into higher bargaining power for the workers. The
positive sign on the cross product between profits and import shares in Table 10 would confirm
this theory. On the other hand, rent sharing seems to be unaffected by an increase in the value of
exports relative to sales.
13
Conclusions
Until now, the economic literature has explored the differences in rent sharing between domestic
and foreign firms, whether they be affiliates or heaquarters of a multinational group. We contribute to this stream of literature by extending the analysis to firms engaged in international
trade, distinguishing between one-way and two-way traders. We formulate the hypothesis that
engaging in trade influences the rent sharing behaviour of firms both by changing the relative
bargaining power of employers and employees, and by increasing the firm’s profitability. We test
our hypotheses on a panel of Slovenian manufacturing firms from 1994 to 2002.
Our estimation takes into account the presence of autocorrelation in the errors, possible omitted
explanatory variables and the endogeneity in the relation between profits and wages. Our estimation confirms the existence of international rent sharing as in Budd et al. (2005), showing that
profit sharing results in a higher wage premium if a firm is either a foreign affiliate or a parent
within an international group. We also find that profit sharing is positive and significant for firms
which are both importing and exporting at the same time, once the change in capital-labour ratio
has been controlled for. If a firm only imports or exports, on the other hand, our evidence is less
clear cut, as the sign of the profit sharing for this kind of firms switches sign once correcting for
endogeneity. Our findings are robust to the introduction of different controls and subsampling
with only negative profits.
14
References
J. M. Abowd and T. Lemieux (1993), “The effects of product market competition on collective
bargaining agreements: The case of foreign competition in Canada”, The Quarterly Journal of
Economics, 108(4), 983–1014.
F. Abraham, J. Konings, and S. Vanormelingen (2009), “The effect of globalization on union
bargaining and price-cost margins of firms”, Review of World Economics (Weltwirtschaftliches
Archiv), 145(1), 13–36.
B. Aitken, A. Harrison, and R. E. Lipsey (1996), “Wages and foreign ownership a comparative
study of Mexico, Venezuela, and the United States”, Journal of International Economics, 40(34), 345–371.
R. Almeida (2007), “The labor market effects of foreign owned firms”, Journal of International
Economics, 72(1), 75–96.
M. Arai and F. Heyman (2009), “Microdata evidence on rent-sharing”, Applied Economics,
41(23), 2965–2976.
M. Arellano and S. Bond (1991), “Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations”, Review of Economic Studies, 58(2), 277–
97.
M. Arellano and O. Bover (1995), “Another look at the instrumental variable estimation of errorcomponents models”, Journal of Econometrics, 68(1), 29–51.
A. B. Bernard and J. B. Jensen (1997), “Exporters, skill upgrading, and the wage gap”, Journal
of International Economics, 42(1-2), 3–31.
D. G. Blanchflower, A. J. Oswald, and P. Sanfey (1996), “Wages, profits, and rent-sharing”, The
Quarterly Journal of Economics, 111(1), 227–51.
R. Blundell and S. Bond (1998), “Initial conditions and moment restrictions in dynamic panel
data models”, Journal of Econometrics, 87(1), 115–143.
E. Brock and S. Dobbelaere (2006), “Has international trade affected workers bargaining
power?”, Review of World Economics (Weltwirtschaftliches Archiv), 142(2), 233–266.
J. W. Budd, J. Konings, and M. J. Slaughter (2005), “Wages and international rent sharing in
multinational firms”, The Review of Economics and Statistics, 87(1), 73–84.
J. W. Budd and M. J. Slaughter (2004), “Are profits shared across borders? evidence on international rent sharing”, Journal of Labor Economics.
D. Card, F. Devicienti, and A. Maida (2010), “Rent-sharing, holdup, and wages: Evidence from
matched panel data”, Nber working papers, National Bureau of Economic Research, Inc.
A. K. Cosar, N. Guner, and J. Tybout (2010), “Firm dynamics, job turnover, and wage distributions in an open economy”, Nber working papers, National Bureau of Economic Research,
Inc.
15
J. Duffy, C. Papageorgiou, and F. Perez-Sebastian (2004), “Capital-skill complementarity? Evidence from a panel of countries”, The Review of Economics and Statistics, 86(1), 327–344.
R. C. Feenstra and G. H. Hanson (1997), “Foreign direct investment and relative wages: Evidence
from Mexico’s Maquiladoras”, Journal of International Economics, 42(3-4), 371–393.
J. A. Frias, D. S. Kaplan, and E. Verhoogen (2012), “Exports and within-plant wage distributions:
Evidence from Mexico”, American Economic Review, 102(3), 435–40.
S. Girma and H. Gorg (2007), “Evaluating the foreign ownership wage premium using a
difference-in-differences matching approach”, Journal of International Economics, 72(1), 97–
112.
P. K. Goldberg and N. Pavcnik (2007), “Distributional effects of globalization in developing
countries”, Journal of Economic Literature, 45(1), 39–82.
Z. Griliches (1969), “Capital-skill complementarity”, The Review of Economics and Statistics,
51(4), 465–68.
Z. Griliches and J. A. Hausman (1986), “Errors in variables in panel data”, Journal of Econometrics, 31(1), 93–118.
G. H. Hanson and A. Harrison (1995), “Trade, technology, and wage inequality”, Nber working
papers, National Bureau of Economic Research, Inc.
E. Helpman, O. Itskhoki, and S. Redding (2010), “Inequality and unemployment in a global
economy”, Econometrica, 78(4), 1239–1283.
E. Helpman, O. Itskhoki, and S. Redding (2011), “Trade and labor market outcomes”, Nber
working papers, National Bureau of Economic Research, Inc.
A. K. Hildreth and A. J. Oswald (1997), “Rent-sharing and wages: Evidence from company and
establishment panels”, Journal of Labor Economics, 15(2), 318–37.
A. Irarrazabal, A. Moxnes, and K.-H. Ulltveit-Moe (2009), “Heterogeneous firms or heterogeneous workers? implications for the exporter premium and the impact of labor reallocation on
productivity”, CEPR discussion papers, Centre for European Policy Research.
F. Kramarz (2003), “Wages and international trade”, CEPR discussion papers, Centre for European Policy Research.
R. A. Lester (1952), “A range theory of wage differentials”, Industrial and Labor Relations
Review, 5(4), 483–500.
J. Levinsohn and A. Petrin (2003), “Estimating production functions using inputs to control for
unobservables”, Review of Economic Studies, 70(2), 317–341.
M. Macis and F. Schivardi (2012), “Exports and wages: Rent sharing, workforce composition or
returns to skills?”, IZA discussion papers, Institute for the Study of Labor (IZA).
D. N. Margolis and K. G. Salvanes (2001), “Do firms really share rents with their workers?”, IZA
discussion papers, Institute for the Study of Labor (IZA).
16
P. S. Martins and Y. Yang (2010), “Globalised labour markets? International rent sharing across
47 countries”, IZA discussion papers, Institute for the Study of Labor (IZA).
J. Matthias Arnold and B. S. Javorcik (2009), “Gifted kids or pushy parents? Foreign direct
investment and plant productivity in Indonesia”, Journal of International Economics, 79(1),
42–53.
D. Rodrik (1997), Has Globalization Gone Too Far?, No. 57 in Peterson Institute Press: All
Books, Peterson Institute for International Economics.
F. Sjholm and R. E. Lipsey (2006), “Foreign firms and Indonesian manufacturing wages: An
analysis with panel data”, Economic Development and Cultural Change, 55.
J. Van Reenen (1996), “The creation and capture of rents: Wages and innovation in a panel of
UK companies”, The Quarterly Journal of Economics, 111(1), 195–226.
E. A. Verhoogen (2008), “Trade, quality upgrading, and wage inequality in the Mexican manufacturing sector”, The Quarterly Journal of Economics, 123(2), 489–530.
F. Windmeijer (2005), “A finite sample correction for the variance of linear efficient two-step
gmm estimators”, Journal of Econometrics, 126(1), 25–51.
M. Yasar and C. J. Morrison Paul (2008), “Capital-skill complementarity, productivity and wages:
Evidence from plant-level data for a developing country”, Labour Economics, 15(1), 1–17.
17
Appendix of tables and figures
Figure 1: Distribution of deflated profits and wages, by firm type.
(b) Profits distribution (Trade)
(a) Profits distribution (Ownership)
.0004
Profits by Trade Status
0
.0001
.0001
Density
.0002
.0003
Density
.0002 .0003
.0004
Profits by Ownership Type
0
−20000
−20000
−10000
0
10000
Average Profits in real Tolars (’000)
Foreign
20000
Export only
Import only
Twoway Trade
Domestic
(c) Wage distribution (Ownership)
(d) Wage distribution (Trade)
Wages by Trade Type
0
.0002
.0002
Density
.0004
.0006
Density
.0004 .0006
.0008
Wages by Ownership Type
.0008
−10000
0
10000
Average Profits in real Tolars (’000)
20000
0
0
0
2000
4000
6000
Average Wage in real Tolars (’000)
Foreign
8000
10000
2000
4000
6000
Average Wage in real Tolars (’000)
Export only
Import only
Twoway Trade
Domestic
18
8000
10000
Figure 2: Time trend for mean deflated profits and wages, by firm type.
(b) Profits development (Trade)
Average Profits by Ownership Type
Average Profits by Trade Type
Domestic
(c) Wage development (Ownership)
02
20
01
20
00
20
99
19
(d) Wage development (Trade)
Average Wages by Ownership Type
Average Wages by Trade Type
02
Domestic
19
02
20
01
20
00
20
99
19
98
19
97
19
96
19
95
Export only
Import only
Twoway Trade
Year
Foreign
19
Year
20
01
20
00
20
99
19
98
19
97
19
96
19
95
19
19
94
19
94
Mean Wage, Real Tolars (’000)
1000 1200 1400 1600 1800
Mean Wage, Real Tolars (’000)
1200 1400 1600 1800 2000 2200
98
Export only
Import only
Twoway Trade
Year
Foreign
19
97
19
96
19
95
19
02
Year
20
01
20
00
20
99
19
98
19
97
19
96
19
95
19
19
94
19
0
94
Mean Profits, Real Tolars (’000)
−10000 5000
20000 35000
Mean Profits, Real Tolars (’000)
50000
100000
150000
(a) Profits development (Ownership)
20
Wage
Π/N
K/N
N
Sales
Value Added
Foreign
1,539.28
4,071.64
32,823.15
191.10
3,620,556
831,759.40
St Dev
2,669.81
1,572.10
783.55
4,636.36
9,496.07
24,612.21
121.70
269.72
2,530,034 11,300,000
583,385.70
1,805,958
2,019.54
304.93
6,889.71
49.96
564,891.40
158,399.40
Mean
1,992
1,995
1,981
1,995
1,995
1,994
37,108
37,262
36,590
37,262
37,262
37,217
n
Wage
Π/N
K/N
N
Sales
Value Added
Wage
Π/N
K/N
N
Sales
Value Added
Export
Wage
1,982.65
1,529.16 35,116
Wage
Π/N
277.86
4,035.72 35,267
Π/N
K/N
6,740.52
33,225.89 34,609
K/N
N
45.90
184.84 35,267
N
Sales 453,726.30
2,531,454 35,267
Sales
Value Added 134,340.60
731,873 35,223
Value Added
Notes: All values are expressed in thousands, except for Employment (units).
Domestic
Wage
Π/N
K/N
N
Sales
Value Added
All Firms
Complete Sample
1,521.77
4,041.68
24,217.54
260.28
5,127,879
1,171,942
St Dev
2,177.79
399.42
7,163.79
89.76
897,686
264,848.40
1,510.74
3,995.99
24,577.61
257.53
3,652,494
1,056,760
2,701.95
1,540.28
774.27
4,420.85
9,410.33
20,576.83
133.86
280.93
2,791,350 11,800,000
643,120.70
1,889,376
2,229.93
436.70
7,387.60
94.15
1,086,059
302,485.70
Mean
16,303
16,322
16,204
16,322
16,322
16,318
1,801
1,803
1,793
1,803
1,803
1,803
18,104
18,125
17,997
18,125
18,125
18,121
n
Import
Wage
Π/N
K/N
N
Sales
Value Added
Wage
Π/N
K/N
N
Sales
Value Added
Wage
Π/N
K/N
N
Sales
Value Added
Table 1: Descriptive statistics by ownership and trade type (a)
1,586.03
4,081.68
29,412.09
236.55
4,626,024
1,059,165
St Dev
2,158.92
1,579.82
445.14
4,020.97
7,253.63 29,817.63
74.12
232.52
746,425.80 3,282,473
220,299.50 949,178.40
2,684.62
1,571.41
793.81
4,656.02
9,573.69
24,724.36
123.11
271.05
2,560,012 11,400,000
590,210.60 1,815,841
2,205.00
475.71
7,457.61
78.42
905,451.40
252,729.10
Mean
20,483
20,507
20,344
20,507
20,507
20,501
1,968
1,971
1,961
1,971
1,971
1,970
22,451
22,478
22,305
22,478
22,478
22,471
n
21
Wage
Π/N
K/N
N
Sales
Value Added
Foreign
1,455.86
-59.42
1,885.82
6.04
68,069.13
23,181.29
1,819.74
247.66
7,013.52
9.00
72,069.21
21,660.85
Mean
1,098.93
2,457.27
2,304.17
13.18
196,978.30
88,129.28
1,356.30
3,205.26
30,073.50
30.86
186,806.00
55,505.46
St Dev
24
24
20
24
24
24
1,859
1,867
1,827
1,867
1,867
1,864
n
Wage
Π/N
K/N
N
Sales
Value Added
Wage
Π/N
K/N
N
Sales
Value Added
Import Only
Wage
1,824.50
1,358.94 1,835
Wage
Π/N
251.66
3,214.17 1,843
Π/N
K/N
7,070.27
30,233.77 1,807
K/N
N
9.03
31.02 1,843
N
Sales 72,121.30 186,725.60 1,843
Sales
Value Added 21,641.02
54,989.95 1,840 Value Added
Notes: All values are expressed in thousands, except for Employment (units).
Domestic
Wage
Π/N
K/N
N
Sales
Value Added
All sample
Export Only
St Dev
n
191
192
188 .
192
192
191
2,005.75
1,698.90 6,015
509.80
3,863.69 6,028
7,442.73 40,906.64 5,947
11.88
37.57 6,028
130,698
491,213 6,028
38,914.80
125,617 6,023
2,366.81
1,822.24
870.68
6,325.78
10,313.84 48,536.03
7.51
19.15
76,110.93
122,049
19,499.98
39,954
2,016.86
1,703.82 6,206
520.94
3,962.33 6,220
7,530.71 41,159.79 6,135
11.74
37.15 6,220
129,013
484,136 6,220
38,318.05
123,914 6,214
Mean
Wage
Π/N
K/N
N
Sales
Value Added
Wage
Π/N
K/N
N
Sales
Value Added
Wage
Π/N
K/N
N
Sales
Value Added
Twoway Trade
Table 2: Descriptive statistics by ownership and trade type (b)
1,532.63
4,126.42
23,464.95
272.92
5,401,787
1,233,086
St Dev
2,222.59
418.22
7,175.52
100.03
1,002,770
295,757.50
1,523.12
4,084.49
23,774.08
271.49
3,864,803
1,117,953
2,718.78
1,538.70
785.52
4,440.74
9,495.20 20,675.63
135.59
282.42
2,828,089 11,900,000
651,484.20 1,900,676
2,276.87
458.41
7,429.87
103.92
1,202,502
334,684.60
Mean
14,468
14,479
14,397
14,479
14,479
14,478
1,777
1,779
1,773
1,779
1,779
1,779
16,245
16,258
16,170
16,258
16,258
16,257
n
Table 3: OLS and FE regressions by foreign and OFDI status
VARIABLES
Π/N
Foreign
Foreign*Π/N
Ofdi
Ofdi*Π/N
K/N
Time dummies
Industry dummies
Firm dummies
1
OLS1
2
OLS2
3
OLS3
4
FE1
5
FE2
6
FE3
0.032*** 0.031*** 0.031*** 0.008*** 0.010***
(0.002)
(0.002)
(0.002)
(0.002)
(0.002)
1.974*** 1.883***
0.206
(0.140)
(0.140)
(0.265)
0.038*** 0.038***
0.039***
(0.009)
(0.009)
(0.010)
2.547***
(0.157)
0.025**
(0.010)
0.010*** 0.010*** 0.010*** 0.012*** 0.012***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
0.011***
(0.002)
0.213
(0.264)
0.032***
(0.010)
0.143
(0.286)
0.074***
(0.010)
0.013***
(0.000)
Yes
Yes
No
Yes
Yes
No
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
EL(base)
EL(foreign)
EL(ofdi)
0.0023
0.0022
0.0049
0.0022
0.0049
0.0040
0.0006
0.0007
0.0035
0.0008
0.0030
0.0061
Lester(base)
Lester(foreign)
Lester(ofdi)
0.1205
0.1166
0.2620
0.1169
0.2605
0.2111
0.0319
0.0373
0.1849
0.0401
0.1612
0.3233
Observations
Number of id
36,458
36,458
36,458
30,585
5,470
30,585
5,470
30,585
5,470
Standard errors in parentheses . *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is
average wage in all models. Foreign is the dummy for Foreign Ownership, Ofdi for Outward
FDI. EL(.) stays for Elasticity of wage to profits per employee, Lester(.) for the Lester Range.
22
Table 4: OLS and FE regressions by trade status only
(1)
OLS
(2)
FE
(3)
OLS
(4)
FE
0.032***
(0.002)
-0.003
(0.085)
-0.005
(0.006)
0.007***
(0.002)
0.025
(0.089)
-0.022***
(0.006)
0.032***
(0.002)
0.009***
(0.002)
0.010***
(0.000)
0.012***
(0.000)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
EL(Base)
EL(Oneway Import)
EL(Oneway Export)
EL(Twoway)
0.0023
0.0019a
0.0005
-0.0010
0.0025
0.0006
0.0025
0.0012
0.00223a
0.0020
0.0047
0.0031
0.0024
0.0031a
0.0040
0.0049
0.0011
-0.0002
0.0029
0.0030
Lester(Base)
Lester(Oneway Import)
Lester(Oneway Export)
Lester(Twoway)
0.1217
0.1016a
0.1323
0.0632
0.2489
0.1652
0.1272
0.1680a
0.2134
0.2629
0.0610
-0.1586
0.0090
0.1568
Observations
Number of id
36,458
36,458
30,585
5,470
36,458
30,585
5,470
VARIABLES
Π/N
Imp
Imp*Π/N
Exp
Exp*Π/N
Both
Both*Π/N
K/L
Time dummies
Industry dummies
Firm dummies
0.0281
-0.0554
30,585
5,470
(5)
OLS
(6)
FE
(7)
OLS
(8)
FE
0.033*** 0.016***
(0.002)
(0.003)
1.362***
0.130
(0.093)
(0.108)
0.011
-0.014**
(0.009)
(0.006)
-0.581***
-0.177
0.789***
-0.063
(0.146)
(0.137)
(0.150)
(0.152)
0.004
0.019*
0.023**
0.025**
(0.011)
(0.011)
(0.011)
(0.011)
1.967***
0.152
2.479***
0.201
(0.065)
(0.101)
(0.074)
(0.130)
0.031*** 0.027*** 0.036*** 0.026***
(0.004)
(0.004)
(0.004)
(0.004)
0.010*** 0.012*** 0.010*** 0.012*** 0.010*** 0.012***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
0.1204
0.0329
0.1344a
0.1070
36,458
30,585
5,470
0.035*** 0.017***
(0.002)
(0.003)
Standard errors in parentheses . *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is average wage in all models. Imp is the
dummy for Importing only, Exp for Exporting only, Both for Twoway. EL(.) stays for Elasticity of wage to profits per employee,
Lester(.) for the Lester Range.“a” indicates elasticities and ranges computed with at least one insignificant coefficient.
23
Table 5: FE regressions, with foreign, OFDI and trading dummies
VARIABLES
Π/N
Foreign
Foreign*Π/N
Ofdi
Ofdi*Π/N
Imp
Imp*Π/N
Exp
Exp*Π/N
Both
Both*Π/N
K/N
1
2
3
4
0.010*** 0.011*** 0.014*** 0.013***
(0.002)
(0.002)
(0.003)
(0.003)
0.204
0.207
0.203
0.183
(0.264)
(0.264)
(0.265)
(0.265)
0.026*** 0.032*** 0.021**
0.018*
(0.010)
(0.010)
(0.010)
(0.010)
0.142
0.139
0.137
0.132
(0.286)
(0.286)
(0.286)
(0.286)
0.075*** 0.075*** 0.063*** 0.065***
(0.010)
(0.010)
(0.011)
(0.011)
0.026
0.125
(0.089)
(0.108)
-0.018***
-0.015**
(0.006)
(0.006)
-0.179
-0.064
(0.137)
(0.152)
0.024**
0.025**
(0.011)
(0.011)
0.166
0.211
(0.101)
(0.130)
0.013*** 0.012**
(0.005)
(0.005)
0.013*** 0.013*** 0.013*** 0.013***
(0.000)
(0.000)
(0.000)
(0.000)
EL(Base)
EL(Foreign)
EL(Ofdi)
EL(Oneway Import)
EL(Oneway Export)
EL(Twoway)
0.0007
0.0025
0.0060
-0.0006
Lester(Base)
Lester(Foreign)
Lester(Ofdi)
Lester(Oneway Import)
Lester(Oneway Export)
Lester(Twoway)
0.0362
0.1352
0.3206
-0.0337
Observations
Number of id
30,585
5,470
0.0008
0.0031
0.0061
0.0010
0.0025
0.0055
0.0025
0.0019
0.0415
0.1652
0.3279
0.0531
0.1342
0.2931
0.1025
0.0504
0.1190
0.2975
-0.0067
0.1455
0.0968
30,585
5,470
30,585
5,470
0.1331
30,585
5,470
0.0009
0.0022
0.0056
-0.0001
0.0027
0.0018
Standard errors in parentheses . *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is average wage in all models. Foreign is the dummy for Foreign
Ownership, Ofdi for Outward FDI, Imp for Importing only, Exp for Exporting only, Both for Twoway. EL(.) stays for Elasticity of wage to profits per
employee, Lester(.) for the Lester Range.“a” indicates elasticities and ranges
computed with at least one insignificant coefficient.
24
Table 6: System GMM estimates
VARIABLES
Π/N
Foreign
Foreign*Π/N
(1)
(2)
-0.001
(0.009)
2.122***
(0.264)
0.071**
(0.030)
0.011**
(0.005)
1.941***
(0.459)
0.041***
(0.015)
1.820***
(0.444)
0.053**
(0.025)
Ofdi
Ofdi*Π/N
Imp
Imp*Π/N
Exp
Exp*Π/N
Both
Both*Π/N
K/N
Time dummies
Industry dummies
Firm dummies
0.005**
(0.003)
(3)
(4)
(5)
0.021***
0.011
0.010
(0.006)
(0.007)
(0.006)
2.052*** 1.934*** 1.319***
(0.535)
(0.412)
(0.486)
0.071*** 0.106* 0.049***
(0.013)
(0.057)
(0.016)
1.614*** 1.396*** 1.119**
(0.396)
(0.375)
(0.486)
0.003
0.061
0.074***
(0.014)
(0.087)
(0.028)
-0.125
(0.830)
0.096***
(0.029)
-2.393*
(1.350)
-0.012
(0.010)
1.865***
(0.564)
0.166***
(0.063)
0.016*** 0.006*** 0.012*** 0.016***
(0.004)
(0.001)
(0.002)
(0.003)
(6)
0.019
(0.012)
1.290***
(0.433)
-0.027
(0.043)
0.694*
(0.409)
-0.034
(0.059)
0.882
(0.981)
0.062**
(0.030)
-1.509
(1.247)
-0.103***
(0.070)
1.930***
(0.468)
0.089***
(0.023)
0.011***
(0.002)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
EL(Base)
EL(Foreign)
EL(Ofdi)
EL(Oneway Import)
EL(Oneway Export)
EL(Twoway)
-0.00005a
0.0045a
0.0008
0.0037
0.0046
0.0015
0.0065
0.0017a
0.0083
0.0008a
0.0083a
0.0051a
0.0007a
0.0042a
0.0060a
0.0013a
-0.0006a
-0.0011a
0.0058a
-0.0104a
0.0077a
Lester(Base)
Lester(Foreign)
Lester(Ofdi)
Lester(Oneway Import)
Lester(Oneway Export)
Lester(Twoway)
-0.0025a
0.2660a
AR(2)
Hansen
Number of IV
Observations
Number of id
-0.0111a
0.0001a
0.0420
0.1962
0.2446
0.0791
0.3491
0.0922a
0.4440
0.0417a
0.4442a
0.2726a
0.0366a
0.2230a
0.3185a
0.0074a
0.0709a
-0.0332a
-0.0581a
0.3081a
-0.2613a
0.4089a
0.909
0.814
18
36,458
5,873
0.947
0.392
84
36,458
5,873
-0.5913a
0.833
0.115
17
36,458
5,873
0.832
0.743
12
36,458
5,873
0.905
0.181
57
36,458
5,873
0.924
0.888
50
36,458
5,873
Standard errors in parentheses . *** p<0.01, ** p<0.05, * p<0.1 . The dependent and independent
variables are defined as above. AR(2) is the p-value of the Arellano Bond test of absence of serial
correlation of order 2. Hansen is the p-value of the Hansen test of joint validity of the instruments.
EL(.) stays for Elasticity of wage to profits per employee, Lester(.) for the Lester Range. “a” indicates
elasticities and ranges computed with at least one insignificant coefficient.
25
26
28,533
5,245
Yes
Yes
Yes
31,178
5,575
Yes
Yes
Yes
-0.013***
(0.001)
-0.003
(0.002)
0.171
(0.264)
0.042***
(0.010)
0.172
(0.288)
0.022***
(0.006)
0.032
(0.089)
-0.006
(0.006)
-0.002
(0.002)
0.042
(0.154)
0.034***
(0.009)
-0.032
(0.162)
0.036***
(0.007)
0.129***
(0.049)
-0.016***
(0.005)
2.622***
(0.047)
2a
1a
28,533
5,245
Yes
Yes
Yes
2.633***
(0.047)
-0.044
(0.076)
0.019*
(0.010)
-0.002
(0.002)
0.042
(0.154)
0.033***
(0.009)
-0.035
(0.162)
0.039***
(0.006)
1b
31,178
5,575
Yes
Yes
Yes
-0.013***
(0.001)
-0.138
(0.137)
0.030***
(0.011)
-0.002
(0.002)
0.165
(0.264)
0.044***
(0.009)
0.167
(0.288)
0.023***
(0.006)
2b
28,533
5,245
Yes
Yes
Yes
-0.021
(0.056)
0.010*
(0.005)
2.624***
(0.047)
-0.004
(0.003)
0.046
(0.154)
0.033***
(0.009)
-0.032
(0.162)
0.034***
(0.007)
1c
31,178
5,575
Yes
Yes
Yes
-0.013***
(0.001)
0.154
(0.101)
0.010**
(0.004)
-0.001
(0.002)
0.165
(0.265)
0.036***
(0.010)
0.163
(0.288)
0.015**
(0.007)
2c
28,533
5,245
Yes
Yes
Yes
0.001
(0.003)
0.036
(0.154)
0.034***
(0.009)
-0.034
(0.162)
0.035***
(0.007)
0.163***
(0.059)
-0.019***
(0.007)
0.010
(0.084)
0.025**
(0.011)
0.073
(0.072)
-0.004
(0.007)
2.622***
(0.047)
1d
31,178
5,575
Yes
Yes
Yes
-0.013***
(0.001)
0.000
(0.002)
0.144
(0.265)
0.036***
(0.010)
0.159
(0.288)
0.016**
(0.007)
0.137
(0.108)
-0.001
(0.006)
-0.018
(0.152)
0.033***
(0.011)
0.224*
(0.130)
0.011***
(0.004)
2d
Standard errors in parentheses . *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is average wage in all models. Foreign
is the dummy for Foreign Ownership, Ofdi for Outward FDI, Imp for Importing only, Exp for Exporting only, Both for Twoway.
TFP is total factor productivity estimated with the LP algorithm, Size the number of employees working in the company.
Observations
Number of id
Time dummies
Industry dummies
Firm dummies
Size
TFP
Both*Π/N
Both
Exp*Π/N
Exp
Imp*Π/N
Imp
Ofdi*Π/N
Ofdi
Foreign*Π/N
Foreign
Π/N
VARIABLES
Table 7: Robustness (1): Alternative controls (FE)
Table 8: Robustness (2): negative profits only (FE)
VARIABLES
Π/N
Foreign
Foreign*Π/N
Ofdi
Ofdi*Π/N
1
-0.020**
(0.008)
1.040***
(0.320)
-0.027
(0.032)
0.973***
(0.334)
0.003
(0.020)
Imp
Imp*Π/N
Exp
Exp*Π/N
Both
Both*Π/N
K/L
Time dummies
Industry dummies
Firm dummies
Observations
Number of id
0.006***
(0.001)
2
3
-0.022*** -0.019***
(0.008)
(0.007)
1.058***
(0.305)
-0.025
(0.029)
0.962***
(0.330)
0.002
(0.019)
0.571***
-0.144
(0.155)
(0.127)
-0.018
-0.018
(0.012)
(0.012)
0.793***
(0.224)
0.019
(0.021)
1.605***
(0.172)
-0.010
(0.014)
0.006*** 0.006***
(0.001)
(0.001)
4
5
6
-0.019** -0.026*** -0.025***
(0.008)
(0.008)
(0.008)
1.024*** 0.848***
0.742**
(0.320)
(0.323)
(0.317)
-0.027
-0.003
-0.006
(0.033)
(0.035)
(0.033)
0.980*** 0.722**
0.673**
(0.336)
(0.333)
(0.327)
0.004
0.023
0.020
(0.021)
(0.025)
(0.024)
0.548***
(0.156)
-0.017
(0.013)
0.034
0.766***
(0.198)
(0.224)
0.022
0.019
(0.020)
(0.021)
1.095*** 1.486***
(0.154)
(0.179)
-0.019
-0.015
(0.017)
(0.017)
0.006*** 0.006*** 0.006***
(0.001)
(0.001)
(0.001)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
12,190
4,354
12,190
4,354
12,190
4,354
12,190
4,354
12,190
4,354
12,190
4,354
Standard errors in parentheses . *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is average
wage in all models. Foreign is the dummy for Foreign Ownership, Ofdi for Outward FDI, Imp for
Importing only, Exp for Exporting only, Both for Twoway.
27
Table 9: Robustness (3): before/after 1997 (FE)
VARIABLES
1
0.014***
(0.003)
Foreign*Π/N
0.013
(0.011)
Ofdi*Π/N
0.064***
(0.011)
Imp*Π/N
-0.022***
(0.007)
Exp*Π/N
0.005
(0.012)
Both*Π/N
0.012**
(0.005)
Π/N*POST
-0.012*
(0.007)
Foreign*Π/N*POST
0.025
(0.030)
Ofdi*Π/N*POST
0.044
(0.064)
Imp*Π/N*POST
0.083***
(0.019)
Exp*Π/N*POST
0.118***
(0.028)
Both*Π/N*POST
0.053***
(0.015)
K/L
0.013***
(0.000)
Π/N
Time dummies
Industry dummies
Firm dummies
Observations
Number of id
2
0.011***
(0.003)
0.016
(0.012)
0.067***
(0.011)
-0.022***
(0.007)
0.015
(0.013)
0.010*
(0.006)
-0.004
(0.006)
0.005
(0.023)
0.001
(0.031)
0.050***
(0.016)
0.051*
(0.026)
0.032**
(0.014)
0.013***
(0.000)
Yes
Yes
Yes
Yes
Yes
Yes
30,585
5,470
30,585
5,470
Standard errors in parentheses . *** p<0.01, **
p<0.05, * p<0.1. The dependent and independent
variables are defined as above, but the coefficients
for the dummies are not reported for presentational
reasons. POST is an indicator for the period after
1997 in column 1, and after 1996 in column 2.
28
Table 10: Robustness (4): intensive margin
VARIABLES
Π/N
Share(Exp)
Share(Exp)*Π/N
1a
OLS
1b
FE
2a
OLS
2b
FE
0.034***
(0.010)
0.191*
(0.110)
0.006
(0.005)
0.014***
(0.002)
0.055
(0.081)
0.003
(0.003)
0.034***
(0.010)
0.017***
(0.003)
Share(Imp)
Share(Imp)*Π/N
K/L
Time dummies
Industry dummies
Firm dummies
Observations
Number of id
0.151*
0.034
(0.078)
(0.089)
0.011** 0.020***
(0.005)
(0.005)
0.011*** 0.013***
(0.002)
(0.001)
3a
OLS
3b
FE
0.034*** 0.017***
(0.010)
(0.003)
0.155
-0.011
(0.124)
(0.094)
0.004
0.001
(0.006)
(0.003)
0.058
0.037
(0.080)
(0.102)
0.007** 0.020***
(0.003)
(0.006)
0.011*** 0.013***
(0.002)
(0.001)
0.011***
(0.002)
0.013***
(0.001)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
36,305
30,440
5,457
36,305
30,440
5,457
36,305
30,440
5,457
Standard errors in parentheses . *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is
average wage in all models. Share(Exp) stays for the share of export sales in the firm’s total sales
revenues, Share(Imp) for that of import.
29